Change
81254f1
81254f1931250a314065cbdada106fe995d83704 · commit on GitHub
pytorch-blog-feed: changed (307331 bytes, HTTP 200)
raw/pytorch-blog-feed/response.xml modified
- Source
- pytorch-blog-feed
- Lines added
- +76
- Lines removed
- -37
- Stored bytes at this commit
- 307,331
- Timestamp
- origin
- Raw artifact at this commit
- raw/pytorch-blog-feed/response.xml
Recorded headers
| observed_at | 2026-11-04T06:22:00.659Z |
|---|---|
| origin_date | 2026-11-04T06:06:14.000Z |
| status | 200 |
| final URL | https://pytorch.org/blog/feed/ |
| etag | "3e30c777cf105a3c4b98fb25b22933dd" |
| last-modified | Tue, 03 Nov 2026 10:15:35 GMT |
| date | Wed, 04 Nov 2026 06:22:00 GMT |
| age | 946 |
| cache-control | public, max-age=60, s-maxage=43200, stale-while-revalidate=86400, stale-if-error=604800 |
| cf-cache-status | null |
| content-encoding | null |
| content-length | 307331 |
@
@@ -12,7 +12,7 @@ <atom:link href="https://pytorch.org/blog/feed/" rel="self" type="application/rss+xml" /> <link>https://pytorch.org/blog/</link> <description></description>-
<lastBuildDate>Mon, 02 Nov 2026 17:13:33 +0000</lastBuildDate>+
<lastBuildDate>Tue, 03 Nov 2026 10:15:35 +0000</lastBuildDate> <language>en-US</language> <sy:updatePeriod> hourly </sy:updatePeriod>@
@@ -28,6 +28,78 @@ <height>32</height></image> <item>+
<title>KVCache Meetup Korea 2026: Event Recap</title>+
<link>https://pytorch.org/blog/kvcache-meetup-korea-2026-event-recap/</link>+
+
<dc:creator><![CDATA[Junghwan Park (PyTorch Foundation Ambassador, Lead maintainer of PyTorchKR)]]></dc:creator>+
<pubDate>Tue, 03 Nov 2026 09:22:03 +0000</pubDate>+
<category><![CDATA[Ambassador Post]]></category>+
<category><![CDATA[Blog]]></category>+
<guid isPermaLink="false">https://pytorch.org/?p=209226</guid>+
+
<description><![CDATA[<p>On the evening of Sep 30, 2026, PyTorchKR and XCENA co-hosted KVCache Meetup Korea. PyTorchKR is a major AI/ML community in Korea, and XCENA is a startup that focuses on...</p>+
<p>The post <a href="https://pytorch.org/blog/kvcache-meetup-korea-2026-event-recap/">KVCache Meetup Korea 2026: Event Recap</a> appeared first on <a href="https://pytorch.org">PyTorch</a>.</p>+
]]></description>+
<content:encoded><![CDATA[<p>On the evening of Sep 30, 2026, <a href="https://pytorch.kr">PyTorchKR</a> and <a href="https://xcena.com">XCENA</a> co-hosted KVCache Meetup Korea. PyTorchKR is a major AI/ML community in Korea, and XCENA is a startup that focuses on CXL (Compute Express Link)…+
<p><em><img fetchpriority="high" decoding="async" class="alignnone wp-image-209227" src="https://pytorch.org/wp-content/uploads/2026/11/01-audience-300x146.jpg" alt="" width="600" height="293" srcset="https://pytorch.org/wp-content/uploads/2026/11/01-audience-300x146.jpg 300w, https://pytorch.org/wp…+
<em>Attendees at KVCache Meetup Korea 2026.</em></p>+
<p>Talks covered CXL memory expansion, GPU and high-bandwidth memory (HBM) bottlenecks, request scheduling on RNGD, and LMCache MP Mode. The sessions showed how hardware and serving software handle KV Cache in different ways. Serving software processes requests that run a model.</p>+
<h2>Registration and Opening</h2>+
<p>The program combined technical talks with opportunities for attendees to connect. Attendees who wanted to join the networking chose up to two interest stickers at registration. The five categories were Model, Serving, Infra, Hardware, and Memory.</p>+
<p><em><img decoding="async" class="alignnone wp-image-209240" src="https://pytorch.org/wp-content/uploads/2026/11/03-registration-300x200.jpg" alt="" width="600" height="400" srcset="https://pytorch.org/wp-content/uploads/2026/11/03-registration-300x200.jpg 300w, https://pytorch.org/wp-content/uplo…+
<em>Attendees chose their interest stickers at registration.</em></p>+
<p>PyTorchKR organizer Junghwan Park and XCENA CPO Juhyun Kim opened the event. The speakers were Jooho Lee from XCENA, Sejin Oh from SK hynix, Donghun Lee from FuriosaAI, and Baoloong Mao from LMCache. They shared their experiences and approaches.</p>+
<p><em><img decoding="async" class="alignnone wp-image-209249" src="https://pytorch.org/wp-content/uploads/2026/11/05-opening-pytorchkr-300x169.jpg" alt="" width="600" height="338" srcset="https://pytorch.org/wp-content/uploads/2026/11/05-opening-pytorchkr-300x169.jpg 300w, https://pytorch.org/wp-co…+
<em>Junghwan Park from PyTorchKR (left) and Juhyun Kim, CPO at XCENA (right), welcomed the attendees.</em></p>+
<h2>CXL and KV Cache Storage: Jooho Lee, XCENA</h2>+
<p>The session “CXL for KVCache” started with the basics of CXL and memory expansion. Jooho Lee explained how CXL expands the memory space that CPUs can access with load/store instructions. He also described a setup where multiple hosts use a memory pool and introduced <a href="https://github.com/xc…+
<p><em><img decoding="async" class="alignnone wp-image-209252 size-large" src="https://pytorch.org/wp-content/uploads/2026/11/07-xcena-1024x683.jpg" alt="" width="1024" height="683" srcset="https://pytorch.org/wp-content/uploads/2026/11/07-xcena-1024x683.jpg 1024w, https://pytorch.org/wp-content/upl…+
<em>Jooho Lee from XCENA presented “CXL for KVCache.”</em></p>+
<p>The KV Cache design combined DRAM and SSD storage in a CXL memory pool. It kept frequently used cache in DRAM and less frequently used cache on SSDs. Maru connects the inference engine to these storage layers. Optimizing LLM inference requires more than storing a large amount of cache. The path a…+
<p>The talk also covered splitting computation for experts in mixture-of-experts (MoE) models between a GPU and MX1. Experts are specialized parts of a model. The design places some weights in memory near MX1 and processes them there. It transfers intermediate values, called activations, and computa…+
<h2>GPU Compute and HBM Bandwidth: Sejin Oh, SK hynix</h2>+
<p>Sejin Oh presented “Analyzing Utilization of GPU and HBM in LLM Serving Systems.” The talk examined how prefill and decode use system resources. Prefill processes the prompt, and decode generates output tokens. The session presented Roofline analysis alongside measurements from a running system. …+
<p><em><img decoding="async" class="alignnone wp-image-209253 size-large" src="https://pytorch.org/wp-content/uploads/2026/11/08-skhynix-session__image-1024x576.png" alt="" width="1024" height="576" srcset="https://pytorch.org/wp-content/uploads/2026/11/08-skhynix-session__image-1024x576.png 1024w, …+
<em>Sejin Oh from SK hynix presented the analysis of GPU compute and HBM bandwidth.</em></p>+
<p>The experiments used two H100 GPUs, Llama 3.1 8B, and vLLM. GPU compute use was high during the main projection operations in prefill. During decode, HBM bandwidth use was high. The experiments also explained the difference between TTFT and TPOT. TTFT measures the time to the first output token. …+
<p>Reusing KV Cache reduces computation during prefill. After that reduction, memory bandwidth can become the bottleneck. Whether a service needs more memory capacity or bandwidth depends on its input and output lengths, batch size, and other operating conditions.</p>+
<h2>Scheduling and KV Cache on RNGD: Donghun Lee, FuriosaAI</h2>+
<p><a href="https://furiosa.ai/">FuriosaAI</a> is building Furiosa-LLM, its own inference engine for running LLMs on RNGD. In “LLM Inference Engine for RNGD,” Donghun Lee shared engineering problems from building the engine, including request scheduling and KV Cache management. He followed each requ…+
<p><em><img decoding="async" class="alignnone wp-image-209254 size-large" src="https://pytorch.org/wp-content/uploads/2026/11/09-furiosaai-1024x576.jpg" alt="" width="1024" height="576" srcset="https://pytorch.org/wp-content/uploads/2026/11/09-furiosaai-1024x576.jpg 1024w, https://pytorch.org/wp-con…+
<em>Donghun Lee from FuriosaAI explained how to share KV Cache outside a node.</em></p>+
<p>Combining requests into one batch increases throughput, or the number of requests processed over time. Larger batches can need larger buckets and increase latency, or response delay. Furiosa-LLM measures the execution cost on the actual hardware. It merges two batches when the combined cost is lo…+
<p>The talk compared the benefit of cache reuse with the cost of transferring it. The slides used a 12,370-token prompt, with one request at a time and three measurements for each condition. Retrieving the host cache and generating the first token took about 92 ms. Recomputing the cache took about 9…+
<p>The benefit depends on the amount of data transferred and the available bandwidth. Experimental features included a network-shared cache and running prefill and decode on separate resources. He also discussed plans for future expansion.</p>+
<h2>‘LMCache: A Unified KV Cache Management Layer with Multi-Platform Architecture’: Baoloong Mao, LMCache</h2>+
<p><a href="https://lmcache.ai/">LMCache</a> maintainer and core architect Baoloong Mao presented “A Unified KV Cache Management Layer with Multi-Platform Architecture.” LMCache is an open-source KV Cache project created by the team that founded <a href="https://tensormesh.ai/about/">Tensormesh</a>.…+
<p><em><img decoding="async" class="alignnone wp-image-209255 size-large" src="https://pytorch.org/wp-content/uploads/2026/11/10-lmcache-1024x576.jpg" alt="" width="1024" height="576" srcset="https://pytorch.org/wp-content/uploads/2026/11/10-lmcache-1024x576.jpg 1024w, https://pytorch.org/wp-content…+
<em>Baoloong Mao from LMCache introduced the unified KV Cache layer and MP Mode.</em></p>+
<p>In MP Mode, CUDA IPC allows processes to share GPU memory. IPC means communication between processes. The processes exchange control information such as block IDs. They use a shared CPU memory pool within a node to reduce duplicate storage. MP Mode also reads and loads data from several storage l…+
<p>The session also covered deployment and operations. Topics included deployment on Kubernetes and monitoring cache hit rates and cache lifetimes. A cache hit means that the requested data is available. The talk also covered isolating failures between inference and cache processes. External plugins…+
<h2>Q&A and Networking</h2>+
<p>In the Q&A session, the speakers returned to the stage to answer questions from attendees. Questions included separating prefill and decode, compared with mixed batching. Mixed batching processes prefill and decode work together. Attendees also asked about the roles of LMCache and the cache f…+
<p><em> <img decoding="async" class="alignnone wp-image-209256" src="https://pytorch.org/wp-content/uploads/2026/11/11-qa-session-300x169.jpg" alt="" width="600" height="338" srcset="https://pytorch.org/wp-content/uploads/2026/11/11-qa-session-300x169.jpg 300w, https://pytorch.org/wp-content/upload…+
<em>The speakers answered live questions together after the talks.</em></p>+
<p>After the Q&A, attendees from different fields gathered in small groups to talk. Their interest stickers helped them find people with similar interests. Speakers and attendees exchanged ideas and heard different perspectives.</p>+
<p>Conversations also continued around the CXL server from XCENA at the venue. Attendees examined the hardware up close and asked questions. They linked the memory expansion techniques from the talks to the hardware on display.</p>+
<p><em><img decoding="async" class="alignnone wp-image-209259" src="https://pytorch.org/wp-content/uploads/2026/11/14-networking-conversation__image-300x200.png" alt="" width="600" height="400" srcset="https://pytorch.org/wp-content/uploads/2026/11/14-networking-conversation__image-300x200.png 300w,…+
<em>Attendees talked during the networking session.</em></p>+
<h2>Feedback from 72 Survey Responses</h2>+
<p>The following summary draws on 72 survey responses. All numbers and percentages below use the survey responses as their base. Some questions allowed multiple selections.</p>+
<p><img decoding="async" class="alignnone wp-image-209261" src="https://pytorch.org/wp-content/uploads/2026/11/19-survey-72-satisfaction-en-300x201.png" alt="" width="600" height="402" srcset="https://pytorch.org/wp-content/uploads/2026/11/19-survey-72-satisfaction-en-300x201.png 300w, https://pytor…+
<em>Event satisfaction and work experience by role, based on 72 survey responses.</em></p>+
<p>In the feedback questions, 48 respondents selected the program as a positive point (66.7%). The venue and facilities received 30 selections (41.7%). The atmosphere for conversations and the event timing each received 22 selections (30.6%). This question allowed multiple selections. Respondents ga…+
<p>Repeated comments mentioned microphone quality and volume, screen visibility from the back rows, and limited space. Some respondents also asked for interpretation or translation to help them follow the English talks. Some respondents also asked for a dedicated session where developers in the same…+
<p><em><img decoding="async" class="alignnone wp-image-209267 size-large" src="https://pytorch.org/wp-content/uploads/2026/11/18-group-photo__image-1024x576.png" alt="" width="1024" height="576" srcset="https://pytorch.org/wp-content/uploads/2026/11/18-group-photo__image-1024x576.png 1024w, https://…+
<em>Group photo from KVCache Meetup Korea.</em></p>+
<p><a href="https://pytorch.kr">PyTorchKR</a> will use this feedback to plan and run future events. Thank you to everyone who joined the event on a weekday evening. The organizers also thank the speakers who shared their knowledge and experience. Thank you to PyTorchKR and XCENA for organizing the e…+
<p>The post <a href="https://pytorch.org/blog/kvcache-meetup-korea-2026-event-recap/">KVCache Meetup Korea 2026: Event Recap</a> appeared first on <a href="https://pytorch.org">PyTorch</a>.</p>+
]]></content:encoded>+
+
+
+
</item>+
<item> <title>Announcing the Special Interest Group (SIG) for Ray on Accelerators</title> <link>https://pytorch.org/blog/announcing-the-special-interest-group-sig-for-ray-on-accelerators/</link> @
@@ -235,7 +307,7 @@<h2><span style="font-weight: 400;">From PyTorch model to CPU instructions</span></h2><p><span style="font-weight: 400;">ExecuTorch is PyTorch’s edge inference framework, with export tools and a lightweight runtime. The export process prepares the model for inference and produces a </span><span style="font-weight: 400;">.pte</span><span style="font-weight: 400;"> file that the device…<p><span style="font-weight: 400;">Supported operations are delegated to </span><a href="https://docs.pytorch.org/executorch/1.3/backends/xnnpack/xnnpack-arch-internals.html"><span style="font-weight: 400;">XNNPACK</span></a><span style="font-weight: 400;">, a CPU inference library. On Android<img s…-
<p><img fetchpriority="high" decoding="async" class="aligncenter wp-image-208006 size-full" src="https://pytorch.org/wp-content/uploads/2026/10/1-2.png" alt="" width="2046" height="1176" srcset="https://pytorch.org/wp-content/uploads/2026/10/1-2.png 2046w, https://pytorch.org/wp-content/uploads/2026…+
<p><img decoding="async" class="aligncenter wp-image-208006 size-full" src="https://pytorch.org/wp-content/uploads/2026/10/1-2.png" alt="" width="2046" height="1176" srcset="https://pytorch.org/wp-content/uploads/2026/10/1-2.png 2046w, https://pytorch.org/wp-content/uploads/2026/10/1-2-300x172.png 3…<p><i><span style="font-weight: 400;">Figure 1. From model export to CPU instructions.</span></i></p><p><span style="font-weight: 400;">SME2 extends Arm’s instruction set to accelerate matrix calculations, including general matrix multiplication (GEMM). In the </span><a href="https://support.arm.com/documentation/107831/0102/The-C1-SME2--unit?lang=en"><span style="font-weight: 400;">Arm C1 CPU clus…<h2><span style="font-weight: 400;">What one streaming call contains</span></h2>@
@@ -509,7 +581,7 @@ in the Agentic Era: Bringing Native PyTorch to Cloud TPUs: Bill Jia (Google Clou <description><![CDATA[<p>The post <a href="https://pytorch.org/blog/meet-the-2026-pytorch-foundation-ambassadors/">Meet the 2026 PyTorch Foundation Ambassadors</a> appeared first on <a href="https://pytorch.org">PyTorch</a>.</p>]]></description> <content:encoded><![CDATA[-
<div id="fws_6ae8cc49a6f9e" data-column-margin="default" data-midnight="dark" class="wpb_row vc_row-fluid vc_row top-level" style="padding-top: 0px; padding-bottom: 0px; "><div class="row-bg-wrap" data-bg-animation="none" data-bg-animation-delay="" data-bg-overlay="false"><div class="inner-wrap…+
<div id="fws_6ae9b449d78ed" data-column-margin="default" data-midnight="dark" class="wpb_row vc_row-fluid vc_row top-level" style="padding-top: 0px; padding-bottom: 0px; "><div class="row-bg-wrap" data-bg-animation="none" data-bg-animation-delay="" data-bg-overlay="false"><div class="inner-wrap… <div class="vc_col-sm-12 wpb_column column_container vc_column_container col no-extra-padding" data-padding-pos="all" data-has-bg-color="false" data-bg-color="" data-bg-opacity="1" data-animation="" data-delay="0" > <div class="vc_column-inner" > <div class="wpb_wrapper">@
@@ -2008,41 +2080,8 @@ Invoke directly on PyTorch tensors (Eager & <code>torch.compile</code>)</td> -
</item>-
<item>-
<title>Nebius Joins the PyTorch Foundation as a Platinum Member</title>-
<link>https://pytorch.org/blog/nebius-joins-the-pytorch-foundation-as-a-platinum-member/</link>-
-
<dc:creator><![CDATA[PyTorch Foundation]]></dc:creator>-
<pubDate>Tue, 20 Oct 2026 16:00:41 +0000</pubDate>-
<category><![CDATA[Announcements]]></category>-
<category><![CDATA[Blog]]></category>-
<category><![CDATA[Press Release]]></category>-
<guid isPermaLink="false">https://pytorch.org/?p=189384</guid>-
-
<description><![CDATA[<p>The PyTorch Foundation, the vendor-neutral home for the open source intelligence layer, is announcing today that Nebius has joined as a Platinum member. Nebius, the AI cloud company, is building...</p>-
<p>The post <a href="https://pytorch.org/blog/nebius-joins-the-pytorch-foundation-as-a-platinum-member/">Nebius Joins the PyTorch Foundation as a Platinum Member</a> appeared first on <a href="https://pytorch.org">PyTorch</a>.</p>-
]]></description>-
<content:encoded><![CDATA[<p><img decoding="async" class="alignnone size-large wp-image-189387" src="https://pytorch.org/wp-content/uploads/2026/10/PyTorch_Foundation_Nebius-1024x576.png" alt="PyTorch Foundation Welcomes Nebius" width="1024" height="576" srcset="https://pytorch.org/wp-cont…-
<p><span style="font-weight: 400;">The PyTorch Foundation, the vendor-neutral home for the open source intelligence layer, is announcing today that Nebius has joined as a Platinum member.</span></p>-
<p><span style="font-weight: 400;">Nebius, the AI cloud company, is building the full-stack platform for developers and companies to take charge of their AI future – from data and model training to production deployment. Founded on deep in-house technological expertise and operating at scale with a …-
<p><span style="font-weight: 400;">Startups and enterprises are using PyTorch to train and fine-tune models at scale on Nebius AI Cloud. Nebius has also long supported vLLM with compute for development, testing, and benchmarking, including multi-node work on large models such as DeepSeek R1. Nebius&…-
<p><span style="font-weight: 400;">“At Nebius, we believe the future of AI depends on a strong, connected ecosystem, from training to deployment. Joining the PyTorch Foundation as a Platinum member allows us to collaborate more deeply with the community and contribute our expertise in high-performan…-
<p><span style="font-weight: 400;">We are excited to welcome Nebius to the PyTorch Foundation as a Platinum Member,” said Mark Collier, Executive Director of the PyTorch Foundation. “The strength of our ecosystem is how we work together to tackle AI’s toughest scaling challenges. B…-
<p><span style="font-weight: 400;">As a platinum member, Nebius is granted one seat to the PyTorch Foundation Governing Board. The Board sets policy through our bylaws, mission and vision statements, describing the overarching scope of foundation initiatives, technical vision, and direction.</span><…-
<p><span style="font-weight: 400;">We’re happy to welcome Ryan Hanrui Wang, Senior Vice President, Applied AI, Nebius, to our board. Ryan Hanrui Wang is SVP, Applied AI at Nebius, where he leads inference, post-training, and agentic systems. He previously co-founded and served as CEO of Eigen …-
<p><span style="font-weight: 400;">To learn more about how your organization can join the PyTorch Foundation, visit our </span><a href="https://pytorch.org/join/"><span style="font-weight: 400;">website</span></a><span style="font-weight: 400;">.</span></p>-
<h2><span style="font-weight: 400;">About PyTorch Foundation</span></h2>-
<p><span style="font-weight: 400;">The PyTorch Foundation is the vendor-neutral home for the open source intelligence layer developers use for training, optimizing, serving, and running models on any chip in any cloud for any agent. As a community-driven hub hosted by the Linux Foundation, the PyTor…-
<h2><span style="font-weight: 400;">About Nebius</span></h2>-
<p><span style="font-weight: 400;">Nebius, the AI cloud company, is building the full-stack platform for developers and companies to take charge of their AI future — from data and model training to production deployment. Founded on deep in-house technological expertise and operating at scale with a …-
<p>The post <a href="https://pytorch.org/blog/nebius-joins-the-pytorch-foundation-as-a-platinum-member/">Nebius Joins the PyTorch Foundation as a Platinum Member</a> appeared first on <a href="https://pytorch.org">PyTorch</a>.</p>-
]]></content:encoded>-
-
-
</item> </channel></rss>-
<!-- plugin=object-cache-pro client=phpredis metric#hits=4747 metric#misses=31 metric#hit-ratio=99.4 metric#bytes=1564895 metric#prefetches=262 metric#store-reads=35 metric#store-writes=8 metric#store-hits=270 metric#store-misses=15 metric#sql-queries=20 metric#ms-total=936.20 metric#ms-cache=43.75 …+
<!-- plugin=object-cache-pro client=phpredis metric#hits=4902 metric#misses=35 metric#hit-ratio=99.3 metric#bytes=1645641 metric#prefetches=295 metric#store-reads=38 metric#store-writes=10 metric#store-hits=304 metric#store-misses=19 metric#sql-queries=22 metric#ms-total=572.99 metric#ms-cache=29.12…45 lines shown here cut at 300 characters. The raw artifact at this commit is linked above.